Distributed Raman fiber amplification has been proven to be a powerful technique to improve the optical signal to noise ratio (OSNR) margin of long haul wavelength-division multiplexing (WDM) system. The discrete Raman fiber amplifier is also an effective method to compensate the loss of the dispersion fiber module and/or provide extra bandwidth. A Raman fiber amplifier can be configured either as a forward-pumped Raman fiber amplifier (RFA) or as a backward-pumped RFA. It has been shown that using both forward-pumped RFA and backward-pumped RFA can achieve better noise performance and Rayleigh crosstalk performance than purely backward pumping, and therefore enables very long span WDM transmission. On the other hand, optical communication is evolving from current point-to-point systems to dynamic optical networks. In a dynamic optical network, channels will be added and dropped to meet the varying capacity demands. In addition, accidental loss of channels due to fiber cut or from amplifier failure will also lead to variation of the overall optical power in the transmission system. To keep the power of the surviving channels at a constant level, fast dynamic gain control is indispensable for both forward-pumped distributed/discrete RFA and backward-pumped distributed/discrete RFA, as well as EDFA's. Two control approaches have been demonstrated in recent years. For the first approach, the Raman pump powers are controlled by a closed negative feedback loop, in which the signal gains are continuously monitored and compared with the target gain. The error control signal is usually generated through a proportional, integral and differential (PID) control algorithm. FIG. 1A shows dynamic gain control apparatus 100 for a multi-wavelength forward-pumped Raman fiber amplifier according to prior art. FIG. 1B shows dynamic gain control apparatus 150 for a multi-wavelength Backward-pumped Raman fiber amplifier according to prior art. This approach exhibits a typical control speed of tens to several hundred microseconds. The corresponding speed may be acceptable for a backward-pumped distributed RFA. This approach is not typically fast enough for a forward-pumped RFA (either distributed or discrete), and many times even not fast enough for a backward-pumped discrete RFA, which typically has much shorter fiber length than a distributed RFA. This observation is due to the fact that the gain transients of a forward-pumped RFA are decided by the walk-off time (sub-us) between the signal and the pump while a backward-pumped RFA is decided by the transit time through the fiber (hundreds of us for a typical distributed RFA).
The second demonstrated method is referred to the all-optical gain clamping technique, which is based on a closed optical feedback loop. However this method introduces noise degradation and is not faster than the first method due to the same nature (closed feedback loop). With another approach, a dynamic gain control scheme based on a predetermined table between the detected output signal power variations and the required pump power adjustments has been proposed for a backward-pumped RFA. Because the look-up table varies with the load (i.e., the power of the input signals), not only is an extra control loop needed to detect the load, but also numerous tables are required to be stored in the control circuits. This not only increases its implementation complexity/cost, but also slows its capability of dynamic gain control.
There is a real need in the art for a fast and efficient dynamic gain control technique suitable for both forward-pumped distributed/discrete RFA and backward-pumped discrete RFA as well as other types of optical fiber amplifiers such as Erbium doped fiber amplifiers (EDFA's).